Six years building a BI practice · Part 5
Leading a BI team through hiring, training and AI
In year six the BI area was built and the work became the team. How we tested candidates, kept training people and let AI shorten learning.
By the sixth year, the BI area I had built inside a consulting firm for private schools in Brazil had all its parts. Data lived in SQL Server, the processing ran on Python, and clients reached their reports through our own web portal. Demand kept growing anyway. The bottleneck had moved, and it was no longer in any system. It was in people.
So my job changed. I built less and led more. More and more of my time went into the team: who joined it, how they learned, and how they kept up with a field that changes every few months.
This is the last part of a series on six years building a BI practice. The previous part covered the client portal with Power BI Embedded.
Two kinds of BI work
When you hire for BI, it helps to admit that the job title covers two different kinds of work.
Operational BI keeps the day running. It means dashboards in tools like Power BI, SQL queries against transactional databases, and knowing how the company's ERP and CRM store their data. The people who do it well solve problems quickly, handle short deadlines without losing their footing, and talk easily with the operations teams who depend on them. Most people start their BI careers here, and it is a good place to learn how a business actually works.
Strategic BI looks further ahead. It uses Python or R, statistics and predictive models, and it needs a working knowledge of business management. The people who do it well see the big picture, find trends in large volumes of data and can explain a complex finding to an executive in plain words.
The move from one to the other is where a BI analyst starts to become a data scientist. At that point, cognitive skills matter more than technical ones. Past a certain level, nobody cares how many languages you know. What matters is how deep the insight is that you pull out of the market you are studying. The analyst stops being someone who builds dashboards and becomes someone who sits next to the client and discusses what to do with the numbers.
How we tested candidates
Hiring is where you shape a team the most, and two tests became the ones I relied on most often. They told me more than any résumé.
The first was a blank table. We handed candidates a dataset with no explanation and asked them to explore it and build the reports they thought were relevant. There was no right answer on file. The test showed whether a person could read raw data, decide what mattered and say why, without being told what to look for.
The second test, also used inside our consulting work, asked for something specific and deliberately outside the person's comfort zone. A typical request: build a geomarketing report with an interactive map that filters every other piece of data on the screen with a click.
The point was that the candidate would not know how to do it at first. That was the test. "I don't know how to do that" was never accepted as an answer, because we were not grading the immediate result. We were watching how the person went after a problem they had never seen.
The analyst I want on a strategic team does not always reach for the most sophisticated tool. Sometimes the clearest solution started on a whiteboard, or on a sheet of paper with a pencil, before anyone opened a laptop.
Training never stops
Hiring well only gets you started. With the area structured, I put real effort into continuous training, so the team kept up with best practices and new technology as the demand from the sector kept rising.
The skills I tried to grow were the same ones I tested for:
- Reading raw data and finding the indicators that matter without detailed instructions.
- Facing an unfamiliar problem with initiative instead of waiting for a recipe.
- Learning a new tool or language when the job needs it, and choosing the simple solution when the simple solution is enough.
- Enough statistics to build and question a predictive model.
- Translating a technical finding into a recommendation a non-technical manager can act on.
That last one is the most underrated. A correct analysis that nobody understands changes nothing.
AI changed the learning curve
One shift deserves its own section, because it changed how I think about training. With AI tools, the time it takes to learn a technical skill has dropped sharply. "I didn't have time to learn it" stopped being a reasonable excuse.
Two consequences follow. First, AI has become the main tool for almost any technical task. Second, a data professional needs to be fluent in learning itself: quick to get past a technical obstacle and to find a better way. The demand chooses the tool, not the analyst's past experience.
It also lightened the operational load. Analysts can now build small custom tools to automate their own repetitive work, and the time they get back goes into analysis, which is the part clients value.
Data culture starts with the leader
A BI team can be excellent and still fail if the organization around it does not use data. That is a leadership problem before it is a technical one.
The practices that matter are simple. Leaders who use data in their own decisions, visibly, set the tone faster than any policy. Clear communication about what the BI work is for builds trust and lowers resistance. Training belongs to everyone who reads the reports, from executives to analysts, not just to the people who build them. And when a decision based on data goes well, it is worth saying so out loud.
None of this replaces the technology. It is what makes the technology get used.
What I would do again
Six years in, these are the choices I would repeat with any new BI team.
- Hire for how people attack problems, not for the list of tools on their résumé. Tools change; the habit of figuring things out does not.
- Test with real, uncomfortable tasks. A blank table and an impossible-looking map reveal more than an interview.
- Make training continuous and part of the job, not a perk for slow weeks.
- Treat AI as the default starting point for learning anything technical, and expect the team to use it.
- Lead by example with data. The team watches what the leader actually does.
The systems from the earlier years made growth possible. The team made it last. If you are structuring a BI area and want to compare notes, write to us.